This is just a vent. Disregard if not interested. Since I started my DeGoogle journey, I
- switched to a more trustworthy VPN that I use more stringently than before;
- deconvoluted personal and work accounts;
- deleted third-party Google logins and set up email / PW logins instead for corporate sites that I do use; I also started using
- NextTube;
- Lemmy;
- Matrix;
- DuckDuckGo, Qwant;
- LibreWolf, IronFox, Mullvad Browser;
- more open source productivity programs, where work admins permit;
- avoid AI use or use locally-hosted or responsibly managed AI where it actually gives me a productivity benefit.
The more I DeGoogle, the more the wonderful rainbow land that should be the internet just starts to feel like a deep dark forest where everything is purposefully designed to exploit the user or wall them off. It’s tiring. User name checks out, I guess. Thanks for listening to my rant, please go on with your -hopefully- wonderful days now.
Consistently using Mullvad Browser alongside a strict VPN is a strong defense against fingerprinting and correlation attacks, but be mindful that the combination can sometimes leak entropy through timing or TLS fingerprinting if not configured carefully. Have you considered whether your local AI setup might inadvertently leak context or model weights to the network if not strictly air-gapped or sandboxed?
I have not. To be perfectly honest, I don’t really understand how that would even work. Can you elaborate? At home, I run a local model with a Kobold CCP backbone to localhost. The physical network is a private Wifi, though the computer is running VPN and I haven’t given much thought about what that means for the AI via localhost. At work, I can thankfully use a responsibly managed AI (company servers, very strong and externally audited data privacy standard with zero on-server data retention) for coding.